BDNF Expression in Larval and Adult Zebrafish Brain: Distribution and Cell Identification

Brain-derived neurotrophic factor (BDNF), a member of the neurotrophin family, has emerged as an active mediator in many essential functions in the central nervous system of mammals. BDNF plays significant roles in neurogenesis, neuronal maturation and/or synaptic plasticity and is involved in cognitive functions such as learning and memory. Despite the vast literature present in mammals, studies devoted to BDNF in the brain of other animal models are scarse. Zebrafish is a teleost fish widely known for developmental genetic studies and is emerging as model for translational neuroscience research. In addition, its brain shows many sites of adult neurogenesis allowing higher regenerative properties after traumatic injuries. To add further knowledge on neurotrophic factors in vertebrate brain models, we decided to determine the distribution of bdnf mRNAs in the larval and adult zebrafish brain and to characterize the phenotype of cells expressing bdnf mRNAs by means of double staining studies. Our results showed that bdnf mRNAs were widely expressed in the brain of 7 days old larvae and throughout the whole brain of mature female and male zebrafish. In adults, bdnf mRNAs were mainly observed in the dorsal telencephalon, preoptic area, dorsal thalamus, posterior tuberculum, hypothalamus, synencephalon, optic tectum and medulla oblongata. By combining immunohistochemistry with in situ hybridization, we showed that bdnf mRNAs were never expressed by radial glial cells or proliferating cells. By contrast, bdnf transcripts were expressed in cells with neuronal phenotype in all brain regions investigated. Our results provide the first demonstration that the brain of zebrafish expresses bdnf mRNAs in neurons and open new fields of research on the role of the BDNF factor in brain mechanisms in normal and brain repairs situations.

Bdnf mRNAs and protein distributions have been described in the brain of different vertebrates species during development and in adults. In rat, bdnf transcripts dramatically increased between embryonic days 11 and 12, a timing coinciding with neurogenesis [20]. Studies performed in rat, mouse and pig showed that bdnf transcripts and proteins were detected at different postnatal stages (newborn, adult) in regions such as hippocampus, hypothalamus, cerebral cortex, amygdala and brain stem adrenergic nuclei [20][21][22][23][24][25][26][27]. In the human brain, bdnf mRNAs were also reported in the hippocampus, the amygdala and the septum [28,29]. There are few studies in birds and amphibians, where BDNF has been described. In songbirds, bdnf is expressed in brain nuclei involved in sensorimotor integration of song learning [30][31][32]. In amphibians, bdnf mRNAs and immunoreactivity were described in optic tectum and hypothalamus [33,34]. In adult teleost fishes, bdnf mRNAs are reported in the brain of eel [35], perciform (Cichlasoma dimerus) [36] and Turquoise killifish (Nothobranchius furzeri) [37]. In zebrafish, bdnf mRNAs are present early during embryonic development and whole-mount in situ hybridization experiments at early developmental stages demonstrated the presence of bdnf transcripts in the forebrain, midbrain and hindbrain [38].
A wealth of data obtained from rodent studies points to a role for BDNF in neurogenesis. BDNF contributes to the regulation of neurogenesis during development by promoting neuronal survival and differentiation [39,40], but also plays roles in adult neurogenesis. However, BDNF effects in adult neurogenesis are not completely established and studies have reported conflicting results. In some studies, BDNF infusion into the ventricle of adult rat substantially increased the number of newly formed cells in many regions, the majority of which differentiate into neurons [41,42]. BDNF administration in the hippocampus was associated with an increased neurogenesis of granule cells in the dentate gyrus [43]. However, more recently, Galvao and colleagues obtained contrasting results suggesting that BDNF delivered intraventricularly in mice and rats failed to enhance subventricular zone neurogenesis and even reduced it [44]. Transgenic mouse models with bdnf gene expression or signalling disruption have been produced to decipher the role of BDNF in regulating neural stem cells but again a clear role of BDNF cannot be stated. Mutant mice lacking BDNF did not survive and had severe neuronal deficits [45,46]. In heterozygous BDNF knockout mice, proliferation of neural stem cells was decreased in the dentate gyrus of the hippocampus [47]. In contrast, conditional knockout mice with depletion of BDNF in mature neurons exhibited an increase in hippocampal proliferation [48]. The role of BDNF in brain regenerative processes was also investigated in rodents. After experimentally induced brain injury, the expression of bdnf was up-regulated close to the lesioned area, probably as part of the repair mechanism following trauma [49][50][51]. In the mouse, BDNF treatment improved long-term survival and maturation of newly formed cells derived from the subcallosal zone after cortex injury [52]. Conversely, in transgenic mice in which bdnf gene expression was abolished in astrocytes, a low number of new generated oligodendrocytes and consequently a larger damaged area were observed after hypoperfusion of the carotid artery [53]. All together, those reports on the role of BDNF in neurogenesis generated conflicting results, which need deeper investigations.
Over the last ten years zebrafish has become an interesting model organism to study the molecular and cell biology of the vertebrate brain [54,55] and has emerged as a model species for translational research in various neuroscience areas, such as depressive disorders [56], neurodegeneration [57], motor neuron disease [58], autism spectrum disorders [59] and Alzheimer's disease [60]. In addition, compared to mammals, the brain of adult zebrafish exhibits a high number of proliferative areas distributed along the ventricles of the telencephalon, diencephalon and mesencephalon [61][62][63][64][65], a feature tightly linked to the persistence of radial glial progenitors [61]. Moreover, as shown by several studies, teleost fishes, including zebrafish, exhibit an outstanding capacity to regenerate after brain injury [66][67][68][69]. The factors supporting this intense neurogenic activity in normal physiological and reparative conditions are not identified to date.
Virtually nothing is known on BDNF functions in the central nervous system of fish. Given the potential role of BDNF in neurogenesis mentioned above and the strong proliferative activity in the brain of adult zebrafish, we have investigated the potential link between this neurotrophin and neurogenesis in zebrafish. First, we studied in detail the distribution of bdnf mRNAs in the brain of zebrafish and determined, by means of double staining studies, the phenotype (proliferating cells, glial or neuronal cells) of bdnf-expressing cells.

Animals and tissue processing
Animals were handled and sacrificed in agreement with the guidelines for the use and care of laboratory animals and in compliance with French and European regulations on animal welfare. Animals used in this study were housed in our zebrafish facilities (INRA LPGP, BIOSIT, Rennes, France, agreement number: B 35-238-6) under standard conditions of photoperiod (14 hours light and 10 hours dark) and temperature (28°C). This project was approved by the local animal care and ethics committee (Comité Rennais d'Ethique en matière d'Expérimentation Animale, Rennes, France), under the number EEA B-35-040. Zebrafish did not receive medical treatment prior or during the experience. No deaths occurred in the facilities before the sacrifice of animals used for in situ hybridization and immunohistochemistry experiments. For studies on larvae, adults were spawned, embryos were collected (n = 20) and kept for 7 days in an incubator in 100 ml glass bottles before being sacrificed in ice water and fixed overnight 4°C in phosphate-buffered saline (PBS; pH 7.4) containing 4% paraformaldehyde (PAF). For studies in adults, four months old females (n = 6) and males (n = 6) were sacrificed by overdose of tricaine methanesulfonate (MS-222, 300mg/l). The sex was then determined by direct examination of the gonads under a binocular stereoscopic microscope. After skull opening, the brain was removed and fixed overnight in PBS-PAF. Both larvae and adult brains were processed for paraffin embedding and microtome sections were mounted on poly-lysine slides.

In situ hybridization
Bdnf expression (ie bdnf mRNAs presence in cells) was investigated using in situ hybridization. For bdnf riboprobes synthesis, we used a pCMV-Sport 6.1 plasmid containing the full-length BDNF cDNA (Unigene DR.132862; Entrez Gene 58118). Antisense and sense riboprobes were generated with DIG RNA Labelling Mix (Roche Diagnostic, Indianapolis, US) by in vitro transcription, using T7 polymerase (Roche-Diagnostic) and SP6 polymerase (Roche-Diagnostic) on plasmid linearized by EcoRI and Not1. To check the specificity of the staining, the sense and antisense riboprobes were always hybridized on adjacent sections. No staining was observed on sections hybridized with the sense riboprobe either in larvae and adults (S1 and S2 Figs) as demonstrated in a previous study in zebrafish [38]. Paraffin sections (7 μm) were deparaffinized with OTTIX and rehydrated through a series of graded ethanol (100-30%). Sections were washed in PBS-NaCl (0.85%) and post-fixed for 20 minutes in PBS-PFA 4%. Tissues were rinsed in PBS and treated for 7 minutes with proteinase K (2 mg/ml) diluted in PBS at 37°C. The reaction was stopped in PBS before post-fixation for 20 minutes in 4% PBS-PAF and the slides were rinsed 10 minutes in PBS and 10 minutes in standard saline citrate (SSC 2x). The sections were incubated overnight at 62.5°C in a moist chamber with the probes (1.5 μg/ ml) diluted in hybridization buffer (formamide 50%; SSC 2X, Denhart 5X, yeast tRNA 50 μ/ml, EDTA 4 mM, dextran sulfate 2.5%). On the following day, slides were rinsed with SCC 2x, SCC 2x/formamide 50%, SSC 0.2x and SSC 0.1x. Next, they were dipped in Tris-HCl/NaCl buffer (100 mM, Tris-HCl pH 7.5, 150 mM NaCl) and washed in the same buffer containing 0.1% Triton and 0.5% of milk powder. Sections were incubated overnight at room temperature with anti-digoxigenin alkaline phosphatase Fab fragments (1:2000, Roche Diagnostic). On the next day, slides were rinsed in Tris-HCl/NaCl buffer and washed three times with Tris-HCl 100 mM (pH 8) containing NaCl (100 mM) and MgCl2 (10mM). The hybridization signal was revealed with the HNPP/Fast-Red detection kit (Roche Diagnostic) for 6 or 12 hours according to the manufacturer's instructions. Sections were washed several times in PBS, mounted in Vectashield medium containing DAPI and coverslipped for microscopic analysis.

Immunohistochemistry
To get further insights in the identification of bdnf-expressing cells (ie cells that express bdnf mRNAs), sections were processed for immunohistochemistry just after in situ hybridization. To stop in situ hybridization reaction, sections were dipped 5 minutes in PBS-PAF 4%, washed in PBS and PBS/Triton (0.2%) to perform immunohistochemistry as follows: sections were incubated overnight at room temperature with different primary antibodies diluted in PBS containing 0.5% milk powder. Sections were washed three times in PBS-Triton 0.2% and incubated with goat anti-rabbit or anti-mouse Alexa Fluor 488 (1:200, Invitrogen, ThermoFisher Scientific, France) for 2 hours. Tissue sections were washed in PBS-Triton 0,2%, and slides were mounted with the Vectashield medium containing DAPI for nuclei conterstaining (Vector Laboratories, Burlingame, CA).
Radial glial cells were identified with a rabbit anti-zebrafish aromatase B (this antibody was raised in our laboratory, 1:200) and with a rabbit anti-mouse BLBP (Brain Lipid Binding Protein, 1:100, Chemicon, Temecula, CA, Cat. No. AB9558). The specificity of these antibodies was previously demonstrated in zebrafish [70,71].
Proliferative cells were visualized with a monoclonal antibody raised against PCNA (Proliferating Cell Nuclear Antigen, 1:100, Dako, France, clone PC10, Cat. no. M0879). The specificity of this antibody for PCNA has been validated in many vertebrate species including zebrafish [61,73].

Microscopy
Sections were observed with an epifluorescence microscope (Olympus Provis, equipped with a DP71 digital camera), an epifluorescence Zeiss (Imager Z1, equipped with the Apotome module) or a confocal microscope Leica SP2. Images were processed with either the Olympus (Cell F ) Zeiss (AxioVision4) or Leica (LCS Lite) software. Micrographs were acquired in TIFF format and adjusted for light and contrast before being assembled on plates using Photoshop CS4. The nomenclature for brain nuclei in adults and developing zebrafish were taken from Wullimann et al, and Mueller and Wullimann, respectively [74,75].

Results
Bdnf mRNAs in the brain of 7 days old larvae We first studied the presence of bdnf mRNAs at early developmental stages. To analyze in detail the distribution of bdnf transcripts in the brain of 7 days old larva, we performed in situ hybridization on series of transverse sections of the whole zebrafish brain. The bdnf sense riboprobe did not generate any signal (S1 Fig, [38]), confirming the specificity of the labeling obtained with the antisense riboprobe. The antisense riboprobe generated a positive labeling in several brain regions, demonstrating the presence of bdnf mRNAs. The Table 1 and Fig 1 summarize the results. From anterior to posterior, bdnf mRNAs were detected in the olfactory rosettes and in the anterior telencephalon (Fig 1A-1D). At this level, positive cells were mostly located in the dorsal and medial parts of the dorsal telencephalon. More caudally, labeled cells were still observed in the dorsal telencephalon, preferentially in the central telencephalon and also in the preoptic area (Fig 1E-1H). Bdnf mRNAs were particularly abundant in the thalamic area (Fig 1I and 1J) and were also consistently observed in the optic tectum, particularly in the periventricular layer (Fig 1G and 1H-1K and 1L). Bdnf mRNAs were also detected at different levels of the midbrain tegmentum, in particular in the reticular formation (Fig 2H-2J).
Attempts to identify the phenotype of these bdnf-expressing cells were made by using double staining with either PCNA, aromatase B or HuC/D as markers of cell proliferation, radial glial cells or neurons respectively [61,76]. Data showed that, in each studied region, cells expressing bdnf mRNAs never corresponded to PCNA-positive cells indicating that bdnf-positive cells are post-mitotic (Fig 2A-2D). Bdnf-expressing cells did not co-express aromatase B attesting that they were not radial glia progenitors (data not shown). Regarding the neuronal marker HuC/D, although the coupling of immunohistochemistry with in situ hybridization resulted in some loss of bdnf transcripts, it appeared that in many regions such as the telencephalon, the optic tectum and the midbrain tegmentum, the signals generated overlapped, suggesting that bdnf-expressing cells may rapidly gained at least a neuronal phenotype ( Fig  2E-2J).
Bdnf mRNAs in the brain of adult zebrafish The distribution of bdnf mRNAs was studied on transverse brain sections of sexually mature males and females from the olfactory bulbs to the medulla oblongata. No labeling was observed with the sense probe (S2 Fig, [38]). The distribution of bdnf mRNAs, revealed by the in situ reaction with the antisense riboprobe, is shown in Table 2 and Figs 3 and 4. The overall pattern of bdnf expression was identical in all animals. In particular, bdnf mRNAs were expressed in the same areas/regions in males and in females. There was no obvious difference between sexes, either in staining intensity and labeled-cells number. For this  (Fig 3A-3C' and 3E). The ventral part of the telencephalon exhibited fewer and weakly labeled cells in the posterior zone and intensely stained cells in the entopedoncular nucleus (Fig 3A and 3B and 3E). In the diencephalon, we observed an intense positive signal in the parvocellular (Fig 3D and 3E), and magnocellular nuclei of the preoptic area and in the entopeduncular and suprachiasmatic nuclei (Fig 3G). Bdnf mRNAs are strongly expressed in cells of the habenula, specifically in its dorsal component (Fig 3H-3I) and in the ventrolateral and ventromedial nuclei of the ventral thalamus (Figs 3F and 3F' and 4F), but the intensity of labeling was weaker than in the dorsal thalamic nuclei (see below). Transcripts  were also abundantly reported in the anterior, dorsal, posterior and central posterior nuclei of the dorsal thalamus (Fig 4B and 4B' and 4G). More ventrally, bdnf mRNAs were expressed in the posterior tuberal nucleus (Fig 4D and 4D' and 4F) and preglomerular nuclei. In the hypothalamus, bdnf transcripts were abundantly expressed in periventricular nucleus of its ventral and dorsal part (Fig 4A and 4A', 4D, 4D', 4F and 4G), in the diffuse and central nuclei of the inferior lobe and in the mammilary body. Between diencephalon and mesencephalon, in the so-called synencephalon, the nucleus of the medial longitudinal fascicle displayed few bdnfpositive cells, whereas bdnf mRNAs were highly expressed in the dorsal and ventral periventricular pretectal nuclei. In the mesencephalon, bdnf mRNAs were also observed in the torus longitudinalis. The periventricular gray zone of the optic tectum exhibited numerous cells expressing bdnf trancripts (Fig 4C and 4C' and 4G). In the tegmentum, bdnf mRNA were observed in the central nucleus of the torus semicircularis and in the interpeduncular nucleus.
In the medulla oblongata, bdnf mRNAs were highly expressed in the secondary gustatory nucleus and in few cells of the vagal lobe. Finally, large bdnf-positive perikarya were seen in the superior and inferior reticular formation (Fig 4E and 4E' and 4H).

Bdnf-expressing cells do not proliferate and are not radial glial cells but neurons
In order to identify the nature of bdnf-expressing cells in the brain of adult zebrafish, we performed bdnf in situ hybridization in combination with immunohistochemistry using antibodies directed against different types of markers (proliferation, neuronal or glial markers), as described above. Bdnf-expressing cells were visualized close to the ventricular surface in many regions such as the telencephalon, the preoptic area, the hypothalamus, the thalamus, and the optic tectum. Previous studies showed that in the brain of adult zebrafish, many proliferating cells were located along the ventricles in those regions [61,63]. To investigate if cells undergoing proliferation could be bdnf-expressing cells, we performed PCNA labeling in combination with bdnf in situ hybridization. As expected, PCNA-labeled cells were positioned along the ventricular cavities in the telencephalon (Fig 5A-5A"'), in the thalamus (Fig 5B and 5B') and in the hypothalamus (Fig 5B and 5B"). In those regions, bdnf-expressing cells were observed very close to PCNA-labeled nuclei, but bdnf mRNAs were never expressed by the proliferating cells (PCNA-positive cells). To test whether some of the bdnf-expressing cells could be radial glial cells, we combined bdnf hybridization with immunohistochemistry against aromatase B or BLBP, two well- established markers of radial glial cells in fish and other vertebrates [61,[76][77][78]. As shown in Fig 6, aromatase B-positive cells never co-expressed bdnf mRNAs in the ventral telencephalon ( Fig 6A and 6A'), preoptic area (Fig 6B and 6B'), entopeduncular nucleus ( Fig 6B and 6B") and thalamus (Fig 6C and 6C'). Analysis of sections from the entire brain using an Apotomeequipped Zeiss microscope did not provide any evidence for expression of bdnf mRNAs in radial glia (Fig 6D-6D"). Similarly, double staining with BLBP antibodies failed to demonstrate any bdnf mRNAs in radial glial cells (Fig 6E-6E"). In contrast, antibodies against neuronal markers (MAP2 and acetylated tubulin) indicated that bdnf mRNAs were expressed in cells with neuronal phenotype in all brain regions investigated. Immunohistochemistry using anti-MAP2 antibody revealed that bdnf mRNAs was exclusively present in neuronal cells as shown in the telencephalon with confocal orthogonal projections (Fig 7A and 7B). Immunohistochemistry using anti-acetylated-tubulin antibody confirmed that bdnf was expressed only in neuronal cells in the telencephalon, habenula (Fig 7C), thalamus (Fig 7D), hypothalamus and optic tectum (Fig 7E and 7E').

Discussion
In the present study, we reported, for the first time, that bdnf mRNAs are consistently expressed in the brain of larval and adult zebrafish. Our results also demonstrated that cells expressing bdnf mRNAs do not correspond to radial glial cells or progenitor cells but rather to neurons.  The present study shows that bdnf mRNAs are widely expressed in the brain of 7 days old larvae, which is in agreement with the data previously obtained by RT-PCR and showing that bdnf expression is maternally inherited, drops down at 1 day post-fertilization (dpf) and then progressively increases from 1 to 8 dpf [38]. Our results bring new information by showing that the strongest signal is observed in the dorsal telencephalon (pallium), the preoptic region, the thalamus and the optic tectum. In general, our expression is in good agreement with what has been shown by previous authors [38]. Indeed, the distribution of bdnf mRNAs was addressed by in toto hybridization and established that bdnf was strongly expressed in the pallium, hypothalamus, posterior tuberculum and optic tectum at 1, 2 and 3 dpf. However, there were no detailed information on the precise structures expressing these messengers. In addition, at 7 dpf, bdnf mRNAs level seems to be lower, possibly due to a problem of access of the probe to the brain [38]. Our results show that in all cases, cells expressing bdnf mRNAs were located in the parenchyma and are identified as neurons. Furthermore, bdnf-expressing cells never exhibited PCNA, aromatase B or BLBP staining, indicating that bdnf mRNAs are not expressed in radial glia and/or progenitors. Although the function of BDNF in the developing brain of zebrafish was not addressed in the present study, these data suggest that this factor could be involved in the differentiation and maintenance of newborn neurons, mainly in the forebrain.
In adult zebrafish, bdnf expression has a similar distribution as in the larvae with the most prominent staining in dorsal telencephalon, preoptic area, dorsal thalamus, posterior tuberculum, hypothalamus, synencephalon and optic tectum. A more diffuse and weaker labelling is detected in other brain regions. The pattern of distribution of bdnf mRNAs is relatively similar to that reported in the European eel and the Turquoise killifish [35,37] with some slight differences regarding the ventral telencephalon, diencephalon, tegmentum and rhombencephalon. Compared to the European eel and the turquoise killifish, bdnf expression in zebrafish is less expressed in the ventral telencephalon. By contrast, the staining in the diencephalon appears more widespread in the zebrafish than in the eel. Some slight differences were seen in other diencephalic nuclei, such as nucleus corticalis and nucleus glomerulosus, which were intensely positive in the killifish and are lacking in the brain of zebrafish. Finally, the tegmentum and rhombencephalon of zebrafish showed fewer bdnf-expressing cells than eel and turquoise killifish. Thus, we can conclude that, in zebrafish, bdnf is expressed mostly in the forebrain and just in few areas of midbrain and hindbrain.
The localization of bdnf transcripts in adult zebrafish is similar in both sexes, despite the fact that sex hormones have been recognized to influence the expression of bdnf through a combination of genomic and epigenetic mechanisms [79,80]. The general idea about the regulation of BDNF (protein and gene activation) by estradiol is that estradiol increases bdnf mRNAs levels. However, if we examine at the literature into more detail, the impacts of estrogens on bdnf mRNAs are complex and fluctuate according to the species, age, brain region and treatment [79][80][81]. Preliminary experiments performed in the laboratory and based on RT-PCR in whole brains failed to show any significant effect of estradiol exposure (10 -7 M to 10 -10 M) on bdnf expression in 7dpf zebrafish larvae.
Because in some regions, bdnf-expressing cells in adults were located close to the ventricular surface, we examined whether such cells could correspond to radial glia cells. Indeed, we and others have shown that in adult fish radial glial cells line entirely the ventricles [61,70,82]. Furthermore, it is now recognized that radial glial cells are progenitors and sustain the constant growth of the brain throughout life [61,83]. To investigate if radial glial cells express or not bdnf transcripts, we used aromatase B and BLBP immunohistochemistry in combination with bdnf in situ hybridization [70,71,84]. Interestingly, we could not detect bdnf mRNAs/aromatase B or bdnf mRNAs/BLBP co-staining, suggesting that radial glial cells do not express bdnf under physiological conditions. However, in mammals, bdnf expression has been documented in glial cells under pathological conditions, in particular after brain lesions or around amyloid plaques [28,85,86]. We previously developed a model of telencephalic lesion in adult zebrafish and showed that proliferation increased rapidly after injury in the parenchyma of the injured telencephalon [68]. Thus, it would be interesting to investigate, in this model, if bdnf expression could be induced in glial cells after brain injuries (as in mammals). Double staining with the cell proliferation marker PCNA failed to show any co-expression of bdnf mRNAs and PCNA in the same cells. However, this does not rule out the possibility that they may express bdnf mRNAs later. A study performed in zebra finch showed that, 15 days after bromodeoxyuridine injection (BrdU, a marker for new born cells), BDNF protein is present in new generated cells labeled with BrdU [87]. In our study, double staining with neuronal makers, notably HuC/D, MAP2 and acetylated-tubulin clearly showed that bdnf -expressing cells have a neuronal phenotype in the brain of adult zebrafish. This finding is consistent with the fact that BDNF is a factor of neuronal origin in mammals [22,24,29]. The neuronal marker HuC/D specifically characterizes early-differentiated neurons [88], whereas the markers MAP2 and acetylatedtubulin characterize both early and late differentiated neurons [89,90]. Therefore, we can assume that bdnf mRNAs could be expressed in the early stages of differentiation processes.
It is interesting to note that the distribution of bdnf mRNAs in the brain of zebrafish is quite similar to that reported in mammals, notably in rat, mouse and pig [21][22][23][24][25][26]. Indeed, strong bdnf expression was reported in the cortex and hippocampus, two pallial structures. Because the telencephalon of fish develops by eversion, it is usually considered that the hippocampus equivalent in zebrafish is the dorsolateral region of the dorsal telencephalon [91][92][93], in which abundant bdnf mRNAs are detected. It is also noteworthy that the central part of the telencephalon, regarded as the presumptive equivalent of the isocortex of mammals, also strongly expresses bdnf [22]. Similarly, the preoptic area and in particular the magnocellular neurons express bdnf trancripts, similar to their mammalian counterparts, the paraventricular and supraoptic nuclei. Other structures exhibiting bdnf mRNAs in both fishes and mammals include the habenula, the thalamic region, the mediobasal hypothalamus and the optic tectum (inferior colliculus). These similarities suggest that BDNF functions are conserved between fishes and mammals and have emerged early in the vertebrate lineage. Studies in Drosophila have recently reported the existence of a neurotrophin-related gene (DNT1) [94]. According to this study, this gene would represent the ancestor of the neurotrophin family that diversified through whole genome duplications. Interestingly, in Drosophila, DNT1 exerts trophic and maintenance functions on neurons.
In conclusion, this study presents a detailed pattern of bdnf mRNAs distribution in the brain of larva and adult zebrafish under normal conditions. We show that bdnf transcripts are most abundant in forebrain regions (that are homolog to bdnf-expressing areas in the mammalian brain). We also bring evidences that bdnf mRNAs are expressed in neurons and not in radial glial cells or progenitor cells. As BDNF contributes to the organization and plasticity of neural network in the intact brain in mammals, further studies in fish should aim at investigating its potential roles in the developing and adult brain in physiological conditions, but also in the case of brain recovery after injuries.

Author Contributions
Conceived and designed the experiments: EP OK LDA CL. Performed the experiments: P. Cacialli MMG P. Coumailleau. Analyzed the data: P. Cacialli OK EP CL. Wrote the paper: EP P. Coumailleau OK CL.